A system and method for artificial intelligence blood data analysis

Abstract

The present invention relates to the field of computer data analysis technology, and in particular to a system and method for artificial intelligence blood data analysis, comprising a data acquisition module, a data synchronization module, a blood data network modeling module, an entropy compression and optimization feature module, and a digital twin modeling module. The present invention achieves intelligent processing of the entire process, from blood data acquisition to recommended medication times, through the collaborative operation of a series of modules, including data acquisition, synchronization, modeling, compression and optimization. This system not only efficiently integrates multimodal blood data to construct an accurate DBN model to explore dynamic causal relationships between data, but also ensures the efficiency and accuracy of data processing through the entropy compression and optimization feature module, and utilizes the digital twin and A3C algorithm model to predict the state of quantum eigenvector values. Ultimately, it accurately recommends medication times and sends them to medical data terminals, effectively improving the efficiency of medical data processing.

Description

Technical Field

[0001] The present invention relates to the technical field of computer data analysis, and in particular to a system and method for artificial intelligence blood data analysis. Background Art

[0002] Blood data contains multimodal information, and these data also exhibit spatiotemporal characteristics. Existing analysis methods often treat these modalities in isolation, failing to effectively integrate multimodal information and account for the spatiotemporal synchrony of the data. This results in an incomplete understanding of blood data and makes it difficult to uncover the underlying physiological and pathological insights.

[0003] A Chinese patent application with publication number CN114662623B discloses a classification method and system for blood samples in coagulation testing based on XGBoost, which belongs to the field of smart medical technology. The system obtains coagulation index data of the blood sample to be tested; inputs the coagulation index data of the blood sample into a pre-trained classification model for blood samples in coagulation testing based on XGBoost, extracts low-level features and high-level features from the coagulation index data of the blood sample according to the preset weights of each coagulation index, and fuses the extracted low-level features and high-level features to obtain the test item features; and uses an XGBoost-based classifier to obtain the coagulation test classification results of the blood sample to be tested based on the test item features. However, this solution still has the problem of difficulty in analyzing and predicting the changing trends of blood in the future, resulting in the inability to recommend medication time based on blood change trends and low efficiency of medication time analysis. Summary of the Invention

[0004] To this end, the present invention provides an artificial intelligence blood data analysis system and method to overcome the problem in the prior art that it is difficult to analyze and predict the changing trends of blood in the future, resulting in the inability to recommend medication time based on blood changing trends and low efficiency of medication time analysis.

[0005] To achieve the above objectives, the present invention provides, on one hand, an artificial intelligence blood data analysis system, comprising:

[0006] A data acquisition module, used to collect blood data;

[0007] A data synchronization module, used to synchronize blood data using a data synchronization method to obtain a multimodal biophysical data set;

[0008] The blood data network modeling module is used to input multimodal biophysical datasets into the pre-built DBN model to obtain the recommended dynamic causal graph output by the DBN model;

[0009] The entropy compression and optimization feature module is used to compress and optimize the recommendation dynamic causal graph through compression optimization methods to obtain quantum feature vector values. This is used to judge the compression effect based on the data compression ratio, calibrate the data synchronization method based on the judgment results, and adjust the judgment process of the compression effect based on the quantum feature similarity.

[0010] The digital twin modeling module is used to input the quantum eigenvector value into the pre-built twin model to obtain the predicted quantum eigenvector value state. It is also used to input the predicted quantum eigenvector value state into the pre-built A3C algorithm model to obtain the recommended medication time, and send the recommended medication time to the medical data terminal.

[0011] Furthermore, the data synchronization module synchronizes the blood data using a data synchronization method to obtain a multimodal biophysical data set, wherein the data synchronization method includes:

[0012] Step A1, interpolating the blood data using a linear interpolation algorithm to obtain interpolated blood data, wherein the interpolated blood data includes an interpolated blood molecular vibration spectrum, interpolated blood pressure, interpolated blood flow rate data, an interpolated cell image, an interpolated blood pH value, and an interpolated oxygen partial pressure;

[0013] Step A2, using the SIFT algorithm to synchronize the interpolated blood molecule vibration spectrum and the interpolated cell image in coordinate space to obtain a synchronized blood molecule vibration spectrum and a synchronized cell image;

[0014] Step A3, using a coordinate synchronization method to synchronize the interpolated blood pH value and the interpolated oxygen partial pressure with the synchronized cell image in coordinate space to obtain the synchronized blood pH value and the synchronized oxygen partial pressure;

[0015] Step A4: Output the synchronized blood molecular vibration spectrum, interpolated blood pressure, interpolated blood flow rate data, synchronized cell image, synchronized blood pH value, and synchronized oxygen partial pressure as a multimodal biophysical dataset to obtain a multimodal biophysical dataset.

[0016] Furthermore, the SIFT algorithm of the data synchronization module in step A2 includes:

[0017] Step A21, detecting extreme points of the interpolated blood molecule vibration spectrum and the interpolated cell image using a Gaussian difference pyramid to obtain stable feature points;

[0018] Step A22, constructing a feature descriptor based on the stable feature points to obtain a vibration spectrum feature descriptor and a cell image feature descriptor;

[0019] Step A23, calculate the distance ratio based on the vibration spectrum feature descriptor and the cell image feature descriptor using the Euclidean nearest neighbor algorithm , get the distance ratio , the distance ratio Compare with the preset distance ratio db0, judge the distance ratio based on the comparison result, and output the feature point attributes based on the judgment result, where:

[0020] when When ≤db0, the data synchronization module determines that the distance ratio is small, and outputs the vibration spectrum feature descriptor and the cell image feature descriptor as matching feature points as feature point attributes;

[0021] when >db0, the data synchronization module determines that the distance ratio is large, and outputs the unmatched feature points between the vibration spectrum feature descriptor and the cell image feature descriptor as feature point attributes;

[0022] Step A24: Output the attribute of the feature point as the feature point of the matching feature point to perform coordinate synchronization, and obtain the synchronized blood molecule vibration spectrum and synchronized cell image.

[0023] Furthermore, the blood data network modeling module constructs a DBN model through a DBN model construction method, and inputs a multimodal biophysical data set into the DBN model to obtain a recommended dynamic causal graph output by the DBN model.

[0024] Furthermore, the entropy compression and optimization feature module compresses and optimizes the recommended dynamic causal graph through a compression optimization method to obtain a quantum feature vector value. The compression optimization method includes:

[0025] Step B1, calculating the single feature entropy of the recommendation dynamic causal graph to obtain the single feature entropy value;

[0026] Step B2, using a greedy algorithm to construct a maximum entropy feature subset based on the single feature entropy value, to obtain a maximum entropy feature subset S;

[0027] Step B3, performing DNA sequence mapping on the maximum entropy feature subset S using a DNA sequence mapping method to obtain a DNA sequence;

[0028] Step B4, use quantum optimization method to perform quantum optimization on the DNA sequence to obtain the quantum eigenvector value .

[0029] Furthermore, the entropy compression and optimization feature module calculates the data compression ratio Ys based on the data size Bs of the recommended dynamic causal graph and the data size By of the DNA sequence, sets Ys=100%×Bs / By, compares the data compression ratio Ys with the preset compression ratio Ys0, judges the compression effect based on the comparison result, and calibrates the data synchronization method based on the judgment result, wherein:

[0030] When Ys≥Ys0, the entropy compression and optimization feature module determines that the compression effect is good and does not correct the data synchronization method;

[0031] When Ys<Ys0, the entropy compression and optimization feature module determines that the compression effect is not good, corrects the data synchronization method, and corrects the preset distance ratio db0 according to the correction coefficient Jz, setting Jz=0.87-0.11×e -0.7×(Ys0-Ys) , e is the base of the natural logarithm, and the corrected preset distance ratio db0J is obtained. Set db0J=db0×Jz, replace the preset distance ratio db0 with the corrected preset distance ratio db0J, and re-compare the distance ratio db with the corrected preset distance ratio db0J.

[0032] Furthermore, the entropy compression and optimization feature module is based on the quantum feature vector value and the target quantum eigenvector value Similarity of quantum features Calculate and set , the quantum feature similarity Similarity to preset Compare and judge the similarity based on the comparison results, and adjust the compression effect judgment process based on the judgment results, where:

[0033] when ≥ When , the entropy compression and optimization feature modules determine that the similarity is similar, and the judgment process of the compression effect is not adjusted;

[0034] when < When the entropy compression and optimization feature modules determine that the similarity is not similar, the judgment process of the compression effect is adjusted by adjusting the coefficient Adjust the preset compression ratio Ys0 and set , get the adjusted preset compression ratio Ys0t, set Ys0t=Ys0× , replace the preset compression ratio Ys0 with the adjusted preset compression ratio Ys0t, and re-compare the data compression ratio Ys with the adjusted preset compression ratio.

[0035] Furthermore, the digital twin modeling module constructs the twin model through the twin model construction method to obtain the twin model, inputs the quantum eigenvector value into the twin model, and obtains the predicted quantum eigenvector value state S(t) output by the twin model.

[0036] Furthermore, the digital twin modeling module constructs the A3C algorithm through the A3C algorithm model construction method to obtain the A3C algorithm model, inputs the predicted quantum eigenvector value state S(t) into the A3C algorithm model, obtains the recommended medication time output by the A3C algorithm model, and sends the recommended medication time to the medical data terminal.

[0037] In another aspect, the present invention further provides a method for artificial intelligence blood data analysis, the method comprising:

[0038] Step S1, collecting blood data;

[0039] Step S2, synchronizing the blood data using a data synchronization method to obtain a multimodal biophysical data set;

[0040] Step S3, inputting the multimodal biophysical dataset into the pre-built DBN model to obtain the recommended dynamic causal graph output by the DBN model;

[0041] Step S4: compress and optimize the recommended dynamic causal graph using a compression optimization method to obtain a quantum feature vector value, judge the compression effect based on the data compression ratio, calibrate the data synchronization method based on the judgment result, and adjust the judgment process of the compression effect based on the quantum feature similarity;

[0042] Step S5: Input the quantum eigenvector value into the pre-built twin model to obtain the predicted quantum eigenvector value state, and also input the predicted quantum eigenvector value state into the pre-built A3C algorithm model to obtain the recommended medication time, and send the recommended medication time to the medical data terminal.

[0043] Compared with the prior art, the beneficial effect of the present invention is that the system collects blood data through the data acquisition module so as to analyze and predict the future trend of blood changes. The system also synchronizes the collected blood data through the data synchronization module to make the blood data highly unified, which is convenient for analyzing and predicting the future trend of blood changes. The system also obtains the recommended dynamic causal graph through the blood data network modeling module, and understands the various feature vectors in the blood data according to the recommended dynamic causal graph so as to compress the blood data. The system also compresses and optimizes the recommended dynamic causal graph through the entropy compression and optimization feature module, reduces the amount of calculation required for analyzing blood data and predicting future blood data, improves the analysis and prediction efficiency of blood data, and thus improves the efficiency of medication time analysis. The digital twin modeling module is also used to build a twin model and an A3C algorithm model to analyze and predict the blood change trend in the next 24 hours, and recommend the medication time based on the predicted blood pressure change trend, thereby improving the analysis efficiency of the medical data analysis terminal for medication time. Through the collaborative work of a series of modules such as data collection, synchronization, modeling, compression optimization, etc., the whole process from blood data collection to recommended medication time is realized. It can not only efficiently integrate multimodal blood data to build an accurate DBN model to explore the dynamic causal relationship between data, but also ensure the efficiency and accuracy of data processing through entropy compression and optimization feature modules, and use digital twins and A3C algorithm models to predict the state of quantum eigenvector values, and finally accurately send the recommended medication time to the medical data terminal, effectively improving the efficiency of medical data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a schematic diagram of the structure of the artificial intelligence blood data analysis system of this embodiment;

[0045] Figure 2 Schematic diagram of the flow of the method for artificial intelligence blood data analysis in this embodiment. DETAILED DESCRIPTION

[0046] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0047] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0048] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0049] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0050] See also Figure 1 , which is a schematic diagram of the structure of the artificial intelligence blood data analysis system of this embodiment, the system includes:

[0051] A data acquisition module, used to collect blood data;

[0052] A data synchronization module, configured to synchronize the blood data using a data synchronization method to obtain a multimodal biophysical data set, wherein the data synchronization module is connected to the data acquisition module;

[0053] A blood data network modeling module, configured to input a multimodal biophysical dataset into a pre-built DBN model to obtain a recommended dynamic causal graph output by the DBN model, wherein the blood data network modeling module is connected to a data synchronization module;

[0054] An entropy compression and optimization feature module is used to compress and optimize the recommended dynamic causal graph through a compression optimization method to obtain a quantum feature vector value, which is used to judge the compression effect based on the data compression ratio, calibrate the data synchronization method based on the judgment result, and adjust the judgment process of the compression effect based on the quantum feature similarity. The entropy compression and optimization feature module is connected to the blood data network modeling module;

[0055] The digital twin modeling module is used to input the quantum eigenvector value into the pre-built twin model to obtain the predicted quantum eigenvector value state, and is also used to input the predicted quantum eigenvector value state into the pre-built A3C algorithm model to obtain the recommended medication time, and send the recommended medication time to the medical data terminal. The digital twin modeling module is connected to the entropy compression and optimization feature module.

[0056] Specifically, the artificial intelligence blood data analysis system is used in a medical data analysis terminal. By predicting the future trend of blood changes, it predicts blood and drug reactions and gives recommended medication time so that the medical data terminal can subsequently analyze the medication data. The system collects blood data through a data acquisition module for subsequent analysis and prediction of future blood change trends. The system also synchronizes the collected blood data through a data synchronization module to make the blood data highly unified, which is convenient for subsequent analysis and prediction of future blood change trends. The system also obtains a recommended dynamic causal graph through a blood data network modeling module, and understands various feature vectors in the blood data based on the recommended dynamic causal graph so that the blood data can be compressed subsequently. The system also compresses and optimizes the recommended dynamic causal graph through an entropy compression and optimization feature module to reduce the time required to analyze blood data and predict future blood data. The system reduces the amount of computing power, improves the analysis and prediction efficiency of blood data, and thus improves the efficiency of medication time analysis. The system also constructs a twin model and an A3C algorithm model through the digital twin modeling module to analyze and predict the blood change trend in the next 24 hours, and recommends medication time based on the predicted blood pressure change trend, thereby improving the analysis efficiency of the medical data analysis terminal for medication time. Through the collaborative work of a series of modules such as data acquisition, synchronization, modeling, compression optimization, etc., the whole process from blood data collection to recommended medication time is realized. Intelligent processing can not only efficiently integrate multimodal blood data to build an accurate DBN model to explore the dynamic causal relationship between data, but also ensure the efficiency and accuracy of data processing through entropy compression and optimization feature modules, and use digital twins and A3C algorithm models to predict the state of quantum eigenvector values, and finally accurately send the recommended medication time to the medical data terminal, effectively improving the efficiency of medical data processing.

[0057] Specifically, the full name of the DBN model in Chinese is Dynamic Bayesian Network Model, and its English name is DynamicBayesian Network.

[0058] Specifically, the data acquisition module is used to collect blood data, and the blood data includes blood molecule vibration spectrum, blood pressure, blood flow rate data, cell image, blood pH value and oxygen partial pressure. The data acquisition module collects the blood molecule vibration spectrum through a terahertz biofield imager. The terahertz biofield imager refers to a device that uses terahertz waves to image the electromagnetic properties of biological tissues or cells, such as Qingyuanfengda QT-TO1000. The blood molecule vibration spectrum refers to the spectrum signal reflecting the vibration energy characteristics of chemical bonds inside molecules obtained when using spectroscopy technology for detection. The data acquisition module collects blood pressure, blood flow rate data and cell images through a microfluidic organ chip. The microfluidic organ chip A chip refers to a bionic microdevice, which uses micromachining technology to construct micron-scale microfluidic channels and porous membrane structures on the chip to simulate the physiological microenvironment of human organs. The blood pressure refers to the pressure exerted by blood on the channel wall in the microfluidic channel. The blood flow rate refers to the fluid volume of blood passing through the microfluidic channel in the microfluidic channel. The cell image refers to a high-resolution image of cell morphology, movement or interaction collected in the microfluidic channel. The data acquisition module collects blood pH and oxygen partial pressure through a quantum sensor. The quantum sensor refers to a high-sensitivity detection device based on the principles of quantum mechanics. The oxygen partial pressure refers to the pressure exerted by oxygen molecules physically dissolved in the blood on the blood vessel wall. The blood pH refers to the quantified value of the acid-base balance state of the blood.

[0059] Specifically, the data acquisition module collects blood data so as to perform intelligent analysis on the blood data later, thereby achieving the effect of intelligent analysis of blood data.

[0060] Specifically, the data synchronization module synchronizes the blood data using a data synchronization method to obtain a multimodal biophysical data set. The data synchronization method includes:

[0061] Step A1, interpolating the blood data using a linear interpolation algorithm to obtain interpolated blood data, wherein the interpolated blood data includes an interpolated blood molecular vibration spectrum, interpolated blood pressure, interpolated blood flow rate data, an interpolated cell image, an interpolated blood pH value, and an interpolated oxygen partial pressure;

[0062] Step A2, using the SIFT algorithm to synchronize the interpolated blood molecule vibration spectrum and the interpolated cell image in coordinate space to obtain a synchronized blood molecule vibration spectrum and a synchronized cell image;

[0063] Step A3, using a coordinate synchronization method to synchronize the interpolated blood pH value and the interpolated oxygen partial pressure with the synchronized cell image in coordinate space to obtain the synchronized blood pH value and the synchronized oxygen partial pressure;

[0064] Step A4: Output the synchronized blood molecular vibration spectrum, interpolated blood pressure, interpolated blood flow rate data, synchronized cell image, synchronized blood pH value, and synchronized oxygen partial pressure as a multimodal biophysical dataset to obtain a multimodal biophysical dataset.

[0065] Specifically, in this embodiment, since the blood pressure, blood flow rate data and cell images are all collected by the microfluidic organ chip, the coordinate space synchronization of the interpolated blood pressure, interpolated blood flow rate data and interpolated cell images is not performed. The linear interpolation algorithm refers to an algorithm that uniformly interpolates blood data collected at different frequencies onto a common time series. This embodiment does not limit the specific implementation method of the linear interpolation algorithm. Those skilled in the art can set it according to actual conditions. For example, the time series of the cell image of the microfluidic chip is used as a reference to obtain a reference time point, and the blood molecule vibration spectrum, blood pressure, blood flow rate data, blood pH value and oxygen partial pressure are calculated by linear interpolation at the reference time point. Set ,in, As the reference time point, is the first original acquisition time point adjacent to the reference time point, is the collection value of the first original collection time point, is the second original acquisition time point adjacent to the reference time point, is the collection value at the second original collection time point, The result after interpolation, the two original acquisition time points adjacent to the reference time point refer to the acquisition time points close to the reference time point during the acquisition process of blood molecule vibration spectrum, blood pressure, blood flow rate data, blood pH value and oxygen partial pressure, the SIFT algorithm refers to a computer vision algorithm that generates feature descriptors by extracting feature points with scale, rotation and brightness invariance in the image. The feature descriptor refers to a statistical description of the local area around the image feature point, which is used to quantify the visual attributes of the feature point for feature matching. The visual attribute refers to an object, scene, image or visual element that can be perceived by the eye and processed and understood by the visual system. The characteristics or properties of the solution, the Chinese full name of the SIFT algorithm is the scale-invariant feature transform algorithm, and its English full name is Scale-Invariant Feature Transform, the coordinate synchronization method refers to a method of synchronizing the spatial coordinates of the interpolated blood pH value and the interpolated oxygen partial pressure with the synchronized cell image with the synchronized cell image as a reference, and the spatial coordinate synchronization refers to the process of unifying the position, direction or motion data in different coordinate systems into the same reference coordinate system through mathematical transformation. This embodiment does not limit the specific implementation method of the coordinate synchronization method, such as setting the coordinates of the quantum sensor for collecting blood pH value and oxygen partial pressure to ( , ), is the horizontal coordinate of the quantum sensor, is the ordinate of the quantum sensor, and the coordinates of the quantum sensor are aligned with the coordinates of the cell image to obtain the aligned quantum sensor coordinates ( , ), To align the horizontal coordinates of the post-quantum sensor, To align the vertical coordinate of the post-quantum sensor, set ,in is the width of the cell image in physical resolution, is the length in physical resolution of the cell image.

[0066] Specifically, the data synchronization module synchronizes the blood data through a data synchronization method to obtain a multimodal biophysical data set with unified timestamps, one coordinate system, and spatial alignment, so as to perform subsequent analysis based on the multimodal biophysical data set with unified timestamps, one coordinate system, and spatial alignment.

[0067] Specifically, the SIFT algorithm of the data synchronization module in step A2 includes:

[0068] Step A21, detecting extreme points of the interpolated blood molecule vibration spectrum and the interpolated cell image using a Gaussian difference pyramid to obtain stable feature points;

[0069] Step A22, constructing a feature descriptor based on the stable feature points to obtain a vibration spectrum feature descriptor and a cell image feature descriptor;

[0070] Step A23, calculate the distance ratio based on the vibration spectrum feature descriptor and the cell image feature descriptor using the Euclidean nearest neighbor algorithm , get the distance ratio , the distance ratio Compare with the preset distance ratio db0, judge the distance ratio based on the comparison result, and output the feature point attributes based on the judgment result, where:

[0071] when When ≤db0, the data synchronization module determines that the distance ratio is small, and outputs the vibration spectrum feature descriptor and the cell image feature descriptor as matching feature points as feature point attributes;

[0072] when >db0, the data synchronization module determines that the distance ratio is large, and outputs the unmatched feature points between the vibration spectrum feature descriptor and the cell image feature descriptor as feature point attributes;

[0073] Step A24: Output the attribute of the feature point as the feature point of the matching feature point to perform coordinate synchronization, and obtain the synchronized blood molecule vibration spectrum and synchronized cell image.

[0074] Specifically, this embodiment does not limit the specific implementation method of detecting the extreme points of the interpolated blood molecule vibration spectrum and the interpolated cell image through the Gaussian difference pyramid. For example, the extreme point D(xp,σ) can be calculated and D(xp,σ)=(G(k×σ)-G(σ))×I(h), where xp represents the image pixel coordinates, σ represents the Gaussian kernel scale parameter, and I(h) refers to the image grayscale value. I(h) includes the image grayscale value of the interpolated cell image and the image grayscale value of the interpolated blood molecule vibration spectrum. I(h) can be obtained through image processing software, and σ=1.6 is set, k is the adjacent scale multiple, and k= , G(σ) represents the Gaussian kernel function. This embodiment does not limit the specific implementation method of constructing the feature descriptor based on the stable feature points, such as constructing the feature descriptor by the descriptor construction method in the SIFT algorithm. The descriptor construction method in the SIFT algorithm refers to the method part of the descriptor construction in the SIFT algorithm. This embodiment does not calculate the distance ratio based on the vibration spectrum feature descriptor and the cell image feature descriptor by the Euclidean distance nearest neighbor algorithm. The specific calculation method is limited, such as each vibration spectrum feature descriptor dTHz in the interpolated blood molecule vibration spectrum and each cell image feature descriptor in the interpolated cell image. When calculating the Euclidean distance OD between , according to the interpolated blood molecule vibration spectrum feature descriptor , interpolated cell image and interpolated blood molecule vibration spectrum feature descriptor The nearest feature descriptor and interpolated cell images and interpolated blood molecule vibration spectrum feature descriptors The second closest feature descriptor Distance ratio Calculate and set , Refers to the interpolated blood molecule vibration spectrum feature descriptor Descriptor of the vibration spectrum of blood molecules in interpolated cell images The nearest feature descriptor The Euclidean distance between It refers to the characteristic descriptor of the blood molecule vibration spectrum after the value Descriptor of the vibration spectrum of blood molecules in interpolated cell images The second closest feature descriptor The Euclidean distance between them, the preset distance ratio refers to a preset value for judging the distance ratio. This embodiment does not limit the specific value of the preset distance ratio. For example, the specific value of the preset distance ratio can be set according to the experience of experts in image data synchronization, and 0.6≤db0≤0.7 is set. This embodiment does not limit the specific implementation method of outputting the feature point attributes as the feature point of the matching feature point for coordinate synchronization. For example, the transformation matrix H can be solved by the matching point, and H= ,in,( , ) is the coordinate of the characteristic point Z1 of the blood molecule vibration spectrum after interpolation, ( , ) is the coordinate Z2 of the matching feature point in the cell image, 、 、 、 、 and is an affine change parameter. This embodiment does not limit the specific value of the affine change parameter. For example, it can be assumed that there are n pairs of matching points ( , , , ), is the horizontal coordinate of the matching point of the cell image after interpolation, is the ordinate of the matching point in the cell image after interpolation, is the abscissa of the interpolated blood molecule vibration spectrum, is the ordinate of the interpolated blood molecule vibration spectrum, and the coordinate error of each pair of matching points is: ,in, 、 Represents the horizontal and vertical coordinate errors, establishes a total error function E, which refers to the sum of the squares of the horizontal and vertical coordinate errors of all matching points, and sets ; Solve the parameters by minimizing the total error function E 、 、 、 、 、 , i={1,2,...,n}, i is the order of matching points.

[0075] Specifically, the data synchronization module achieves high-precision cross-modal registration of blood molecule vibration spectra and cell images through robust feature matching and spatial coordinate synchronization, providing a technical basis for multi-scale biophysical information fusion.

[0076] Specifically, the blood data network modeling module constructs a DBN model through a DBN model construction method, and inputs a multimodal biophysical dataset into the DBN model to obtain a recommended dynamic causal graph output by the DBN model.

[0077] Specifically, the DBN model refers to a probabilistic graphical model for modeling time series data that takes a multimodal biophysical data set as input and outputs a recommended dynamic causal graph. The dynamic causal relationship between variables is described by a directed acyclic graph. This embodiment does not limit the specific implementation method of constructing the DBN model through the DBN model construction method. For example, the nodes of the DBN can be determined based on biological prior knowledge, including biological entities such as blood molecules, blood pressure, blood flow rate, cells, blood pH value and oxygen partial pressure; and the causal relationship between the nodes is established through prior knowledge, including a physical causal chain, a molecular-cellular causal chain and a biochemical causal chain. The physical causal chain includes: blood pressure → flow rate → molecular transmission efficiency → cellular microenvironment; the biochemical causal chain includes: oxygen partial pressure → red blood cell oxygen carrying capacity → cellular anaerobic metabolism → lactate accumulation → pH, and the molecular-cellular causal chain includes: inflammatory factors → leukocyte activation → vascular endothelial damage.

[0078] Specifically, the blood data network modeling module builds a DBN model through the DBN model construction method, inputs the multimodal biophysical data set into it to obtain a recommended dynamic causal graph, and can deeply explore the complex dynamic correlation and causal logic between multimodal information in the blood data, so as to subsequently predict the blood change trend.

[0079] Specifically, the entropy compression and optimization feature module compresses and optimizes the recommended dynamic causal graph through a compression optimization method to obtain a quantum feature vector value. The compression optimization method includes:

[0080] Step B1, calculating the single feature entropy of the recommendation dynamic causal graph to obtain the single feature entropy value;

[0081] Step B2, using a greedy algorithm to construct a maximum entropy feature subset based on the single feature entropy value, to obtain a maximum entropy feature subset S;

[0082] Step B3, performing DNA sequence mapping on the maximum entropy feature subset S using a DNA sequence mapping method to obtain a DNA sequence;

[0083] Step B4, use quantum optimization method to perform quantum optimization on the DNA sequence to obtain the quantum eigenvector value .

[0084] Specifically, this embodiment does not limit the specific implementation method of calculating the single feature entropy of the recommended dynamic causal graph. For example, the feature vector of the recommended dynamic causal graph can be calculated based on the feature vector of the recommended dynamic causal graph. ∈{Xh, Xw, Xg, Xs, Xq, Xy}, Recommend dynamic causal graph feature vectors for the causal weight of each feature in the dynamic causal graph Medium status value and recommended dynamic causal graph feature vector Medium status value The probability density of , for single feature entropy Calculate and set This embodiment does not limit the specific numerical value of the causal weight of each feature in the dynamic causal graph. For example, it can be set based on expert analysis of clinical data. The clinical data refers to a multi-dimensional information set related to the patient's health status and treatment process that is systematically collected during medical activities. The Xh refers to the cell microenvironment feature in the recommended dynamic causal graph, Xw refers to the anaerobic metabolism marker feature in the recommended dynamic causal graph, Xg refers to the vascular intradermal loss feature in the recommended dynamic causal graph, Xs refers to the tissue oxygenation index feature in the recommended dynamic causal graph, Xq refers to the inflammation-ischemia score feature in the recommended dynamic causal graph, and Xy refers to the stress time feature in the recommended dynamic causal graph. The feature vector Medium status value Refers to the order of the blood data in the recommended dynamic causal graph, which includes Xh, Xw, Xg, Xs, Xq, Xy, and the feature vector Medium status value The probability density of Refers to the probability of the order of the states of the blood data in the recommended dynamic causal graph. This embodiment does not limit the specific implementation method of using the greedy algorithm to construct the maximum entropy feature subset according to the single feature entropy value. Those skilled in the art can set it according to actual needs. For example, the maximum entropy feature subset S is initialized first, S is set to the empty set, and all single feature entropies are calculated. , select single feature entropy Greater than the preset characteristic entropy 0 feature is added to S, and set 0.8≤ 0≤0.9. This embodiment does not limit the specific implementation method of using the DNA sequence mapping method to map the maximum entropy feature subset S to the DNA sequence. For example, the features in the maximum entropy feature subset S can be normalized to the [0-1] interval to obtain the normalized maximum entropy feature subset S, and the [0,1] interval is divided into 4 equal probability intervals, corresponding to 4 bases, the 4 bases including A, T, C and G, wherein: [0,0.25)→A, [0.25,0.5)→T, [0.5,0.75)→C, [0.75,1]→G, and the normalized maximum entropy feature subset S corresponding to the 4 bases is classified to obtain a DNA sequence, according to A / T→ Mapping A / T in DNA sequences to ground states , according to C / G→ Mapping C / G in DNA sequences to ground states The Chinese full name of the DNA is deoxyribonucleic acid, and its English full name is Deoxyribonucleic Acid.

[0085] Specifically, the entropy compression and optimization feature module retains the uncertainty of biological mechanisms through information entropy theory, combines causal weights to strengthen key pathological associations, and ultimately utilizes the high storage density of DNA and the parallelism of quantum computing to achieve lightweight and high-information-density storage of the feature library.

[0086] Specifically, the entropy compression and optimization feature module calculates the data compression ratio Ys based on the data size Bs of the recommended dynamic causal graph and the data size By of the DNA sequence, sets Ys=100%×Bs / By, compares the data compression ratio Ys with the preset compression ratio Ys0, judges the compression effect based on the comparison result, and calibrates the data synchronization method based on the judgment result, wherein:

[0087] When Ys≥Ys0, the entropy compression and optimization feature module determines that the compression effect is good and does not correct the data synchronization method;

[0088] When Ys<Ys0, the entropy compression and optimization feature module determines that the compression effect is not good, corrects the data synchronization method, and corrects the preset distance ratio db0 according to the correction coefficient Jz, setting Jz=0.87+0.11×e -0.7×(Ys0-Ys) , e is the base of the natural logarithm, and the corrected preset distance ratio db0J is obtained. Set db0J=db0×Jz, replace the preset distance ratio db0 with the corrected preset distance ratio db0J, and re-compare the distance ratio db with the corrected preset distance ratio db0J.

[0089] Specifically, the data size of the recommended dynamic causal graph refers to the information data size of the recommended dynamic causal graph, and the data size of the DNA sequence refers to the information data size of the DNA sequence. This embodiment does not limit the method of obtaining the data size of the recommended dynamic causal graph and the data size of the DNA sequence. For example, the data size of the recommended dynamic causal graph and the data size of the DNA sequence can be obtained through computer background data. The preset compression ratio refers to a preset value for judging the compression effect. This embodiment does not limit the specific value of the preset compression ratio. Those skilled in the art can set it according to actual conditions. For example, according to the analysis efficiency requirements for blood data, when the analysis efficiency requirements are high, 90%≤Ys0≤95% is set. The compression effect refers to the quality of the compression effect of the recommended causal relationship graph judged based on the data compression ratio and the preset compression ratio. The compression effect includes a good compression effect and a poor compression effect.

[0090] Specifically, the entropy compression and optimization feature module judges the compression effect, corrects the data synchronization method, sets the correction coefficient from 0.98 to 0.87 as the data compression ratio decreases, adjusts the preset distance ratio, and reduces the value of the preset distance ratio to increase the matching feature points and make its synchronization accuracy higher.

[0091] Specifically, the entropy compression and optimization feature module is based on the quantum feature vector value and the target quantum eigenvector value Similarity of quantum features Calculate and set , the quantum feature similarity Similarity to preset Compare and judge the similarity based on the comparison results, and adjust the compression effect judgment process based on the judgment results, where:

[0092] when ≥ When , the entropy compression and optimization feature modules determine that the similarity is similar, and the judgment process of the compression effect is not adjusted;

[0093] when < When the entropy compression and optimization feature modules determine that the similarity is not similar, the judgment process of the compression effect is adjusted by adjusting the coefficient Adjust the preset compression ratio Ys0 and set , e is the base of the natural logarithm, and the adjusted preset compression ratio Ys0t is obtained, and Ys0t is set to Ys0× , replace the preset compression ratio Ys0 with the adjusted preset compression ratio Ys0t, and re-compare the data compression ratio Ys with the adjusted preset compression ratio.

[0094] Specifically, the target quantum eigenvector value refers to the quantum eigenvector value that the optimized DNA sequence needs to achieve. This embodiment does not limit the specific numerical value of the target quantum eigenvector value. Those skilled in the art can set it according to their own needs, such as setting the specific numerical value of the target quantum eigenvector value according to the computational efficiency requirement. The preset similarity refers to a preset value for judging the similarity situation. This embodiment does not limit the specific numerical value of the preset similarity. For example, according to the requirement of high information density, 0.9≤Ftx0≤1 is set. The similarity situation refers to the similarity between the quantum eigenvector value judged based on the quantum feature similarity and the preset similarity and the target quantum eigenvector value. The similarity situation includes the similarity situation being similar and the similarity situation being dissimilar.

[0095] Specifically, the entropy compression and optimization feature module judges the similarity and adjusts the compression effect judgment process based on the judgment result to improve the subsequent analysis efficiency of blood data. By setting the adjustment coefficient from 0.99 to 0.85 as the quantum feature similarity decreases, the preset compression ratio value is reduced, thereby reducing the compression ratio and increasing the similarity of the data, avoiding inaccurate analysis of the subsequent blood data and improving the accuracy of blood data analysis.

[0096] Specifically, the digital twin modeling module constructs the twin model through the twin model construction method to obtain the twin model, inputs the quantum eigenvector value into the twin model, and obtains the predicted quantum eigenvector value state S(t) output by the twin model.

[0097] Specifically, the twin model refers to a machine learning model that takes quantum eigenvector values ​​as input and predicts quantum eigenvector value states as output. The predicted quantum eigenvector value state S(t) refers to the quantum eigenvector value state at time t. This embodiment does not limit the specific implementation method of the twin model construction method. Those skilled in the art can set it according to actual conditions, such as according to the state transfer matrix Aj, the drug dosage vector u(t), the drug action matrix Bj, and the random noise. (t) Set the state transfer equation S(t+1) and set , where: S(t)=[s1(t),s2(t),...,sd(t)], S(t) refers to the total set of quantum eigenvector states at time t, s1(t) refers to the quantum eigenvector state at the first time, s2(t) refers to the quantum eigenvector state at the second time, sd(t) refers to the quantum eigenvector state at the dth time, d is the predicted node order of the quantum eigenvector state at time t, and S(t+1) refers to the total set of quantum eigenvector states at time t+1. This embodiment does not limit the method for obtaining the state transfer matrix and the drug action matrix. For example, the state transfer matrix and the drug action matrix can be obtained through statistical fitting of clinical data. This embodiment does not limit the method for obtaining the dosage vector. For example, the dosage vector can be obtained by converting the drug dosage in the doctor's recommendation into a vector form. This embodiment does not limit the method for obtaining random noise. For example, the random noise can be determined based on expert experience. The clinical data refers to a multi-dimensional information set related to the patient's health status and treatment process that is systematically collected during medical activities. The doctor's recommendation refers to the dosage recommendation given by the doctor based on the patient's condition.

[0098] Specifically, the digital twin modeling module inputs the quantum eigenvector value into the twin model to obtain the predicted quantum eigenvector value state S(t) output by the twin model, and predicts the future change trend of the blood data so as to make subsequent recommendations on the medication time.

[0099] Specifically, the digital twin modeling module constructs the A3C algorithm through the A3C algorithm model construction method to obtain the A3C algorithm model, inputs the predicted quantum eigenvector value state S(t) into the A3C algorithm model, obtains the recommended medication time output by the A3C algorithm model, and sends the recommended medication time to the medical data terminal.

[0100] Specifically, the A3C algorithm model refers to a machine learning model that takes the predicted quantum eigenvector value state S(t) as input and the recommended medication time as output. This embodiment does not limit the specific implementation method of the A3C algorithm model construction method. For example, the A3C algorithm can be used as the basic framework of the A3C algorithm model to construct the A3C algorithm model. The A3C algorithm refers to a distributed reinforcement learning framework based on policy gradients, which achieves efficient exploration and policy optimization through asynchronous parallel training. This embodiment does not limit the specific implementation method of sending the recommended medication time to the medical data terminal. For example, the recommended medication time can be sent to the medical data terminal via a wireless signal. The medical data terminal refers to a terminal that analyzes and processes medical data.

[0101] Specifically, the digital twin modeling module outputs the recommended medication time and sends the recommended medication time to the medical data terminal, thereby reducing the computing load of the medical data terminal and effectively improving the efficiency of medical data processing.

[0102] See also Figure 2 , which is a flow chart of the method for artificial intelligence blood data analysis in this embodiment, the method includes:

[0103] Step S1, collecting blood data;

[0104] Step S2, synchronizing the blood data using a data synchronization method to obtain a multimodal biophysical data set;

[0105] Step S3, inputting the multimodal biophysical dataset into the pre-built DBN model to obtain the recommended dynamic causal graph output by the DBN model;

[0106] Step S4: compress and optimize the recommended dynamic causal graph using a compression optimization method to obtain a quantum feature vector value, judge the compression effect based on the data compression ratio, calibrate the data synchronization method based on the judgment result, and adjust the judgment process of the compression effect based on the quantum feature similarity;

[0107] Step S5: Input the quantum eigenvector value into the pre-built twin model to obtain the predicted quantum eigenvector value state, and also input the predicted quantum eigenvector value state into the pre-built A3C algorithm model to obtain the recommended medication time, and send the recommended medication time to the medical data terminal.

[0108] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. An artificial intelligence blood data analysis system, characterized in that: include: A data acquisition module, used to collect blood data; A data synchronization module, used to synchronize blood data using a data synchronization method to obtain a multimodal biophysical data set; The blood data network modeling module is used to input multimodal biophysical datasets into the pre-built DBN model to obtain the recommended dynamic causal graph output by the DBN model; The entropy compression and optimization feature module is used to compress and optimize the recommendation dynamic causal graph through compression optimization methods to obtain quantum feature vector values. This is used to judge the compression effect based on the data compression ratio, calibrate the data synchronization method based on the judgment results, and adjust the judgment process of the compression effect based on the quantum feature similarity. A digital twin modeling module is used to input the quantum eigenvector value into a pre-built twin model to obtain a predicted quantum eigenvector value state. The module is also used to input the predicted quantum eigenvector value state into a pre-built A3C algorithm model to obtain a recommended medication time, and then send the recommended medication time to the medical data terminal. The data synchronization module synchronizes the blood data using a data synchronization method to obtain a multimodal biophysical data set, wherein the data synchronization method includes: Step A1, interpolating the blood data using a linear interpolation algorithm to obtain interpolated blood data, wherein the interpolated blood data includes an interpolated blood molecular vibration spectrum, interpolated blood pressure, interpolated blood flow rate data, an interpolated cell image, an interpolated blood pH value, and an interpolated oxygen partial pressure; Step A2, using the SIFT algorithm to synchronize the interpolated blood molecule vibration spectrum and the interpolated cell image in coordinate space to obtain a synchronized blood molecule vibration spectrum and a synchronized cell image; Step A3, using a coordinate synchronization method to synchronize the interpolated blood pH value and the interpolated oxygen partial pressure with the synchronized cell image in coordinate space to obtain the synchronized blood pH value and the synchronized oxygen partial pressure; Step A4, outputting the synchronized blood molecular vibration spectrum, interpolated blood pressure, interpolated blood flow velocity data, synchronized cell image, synchronized blood pH value, and synchronized oxygen partial pressure as a multimodal biophysical dataset to obtain a multimodal biophysical dataset; The SIFT algorithm of the data synchronization module in step A2 includes: Step A21, detecting extreme points of the interpolated blood molecule vibration spectrum and the interpolated cell image using a Gaussian difference pyramid to obtain stable feature points; Step A22, constructing a feature descriptor based on the stable feature points to obtain a vibration spectrum feature descriptor and a cell image feature descriptor; Step A23, calculate the distance ratio based on the vibration spectrum feature descriptor and the cell image feature descriptor using the Euclidean nearest neighbor algorithm , get the distance ratio , compare the distance ratio db with the preset distance ratio db0, judge the distance ratio according to the comparison result, and output the feature point attributes according to the judgment result, where: when When ≤db0, the data synchronization module determines that the distance ratio is small, and outputs the vibration spectrum feature descriptor and the cell image feature descriptor as matching feature points as feature point attributes; when >db0, the data synchronization module determines that the distance ratio is large, and outputs the unmatched feature points between the vibration spectrum feature descriptor and the cell image feature descriptor as feature point attributes; Step A24, outputting the feature point attributes as the feature point matching the feature point coordinate synchronization, obtaining the synchronized blood molecule vibration spectrum and synchronized cell image; The blood data network modeling module constructs a DBN model using a DBN model construction method, and inputs a multimodal biophysical dataset into the DBN model to obtain a recommended dynamic causal graph output by the DBN model; The entropy compression and optimization feature module compresses and optimizes the recommended dynamic causal graph through a compression optimization method to obtain a quantum feature vector value. The compression optimization method includes: Step B1, calculating the single feature entropy of the recommendation dynamic causal graph to obtain the single feature entropy value; Step B2, using a greedy algorithm to construct a maximum entropy feature subset based on the single feature entropy value, to obtain a maximum entropy feature subset S; Step B3, performing DNA sequence mapping on the maximum entropy feature subset S using a DNA sequence mapping method to obtain a DNA sequence; Step B4, use quantum optimization method to perform quantum optimization on the DNA sequence to obtain the quantum eigenvector value ; The entropy compression and optimization feature module calculates the data compression ratio Ys based on the data size Bs of the recommended dynamic causal graph and the data size By of the DNA sequence, sets Ys=100%×Bs / By, compares the data compression ratio Ys with the preset compression ratio Ys0, judges the compression effect based on the comparison result, and calibrates the data synchronization method based on the judgment result, wherein: When Ys≥Ys0, the entropy compression and optimization feature module determines that the compression effect is good and does not correct the data synchronization method; When Ys<Ys0, the entropy compression and optimization feature module determines that the compression effect is not good, corrects the data synchronization method, and corrects the preset distance ratio db0 according to the correction coefficient Jz, setting Jz=0.87-0.11×e -0.7×(Ys0-Ys) , e is the base of the natural logarithm, get the preset distance ratio db0J after correction, set db0J=db0×Jz, replace the preset distance ratio db0 with the preset distance ratio db0J after correction, and set the distance ratio Re-compare with the calibrated preset distance ratio db0J.

2. The artificial intelligence blood data analysis system according to claim 1, characterized in that: The entropy compression and optimization feature module is based on the quantum feature vector value and the target quantum eigenvector value Similarity of quantum features Calculate and set , the quantum feature similarity Similarity to preset Compare and judge the similarity based on the comparison results, and adjust the compression effect judgment process based on the judgment results, where: when ≥ When , the entropy compression and optimization feature modules determine that the similarity is similar, and the judgment process of the compression effect is not adjusted; when < When the entropy compression and optimization feature modules determine that the similarity is not similar, the judgment process of the compression effect is adjusted by adjusting the coefficient Adjust the preset compression ratio Ys0 and set , get the adjusted preset compression ratio Ys0t, set Ys0t=Ys0× , replace the preset compression ratio Ys0 with the adjusted preset compression ratio Ys0t, and re-compare the data compression ratio Ys with the adjusted preset compression ratio.

3. The artificial intelligence blood data analysis system according to claim 2, characterized in that: The digital twin modeling module constructs the twin model through the twin model construction method to obtain the twin model, inputs the quantum eigenvector value into the twin model, and obtains the predicted quantum eigenvector value state S(t) output by the twin model.

4. The artificial intelligence blood data analysis system according to claim 3, characterized in that: The digital twin modeling module constructs the A3C algorithm through the A3C algorithm model construction method to obtain the A3C algorithm model, inputs the predicted quantum eigenvector value state S(t) into the A3C algorithm model, obtains the recommended medication time output by the A3C algorithm model, and sends the recommended medication time to the medical data terminal.

Images